Case study

LLM Security Log Assistant

A self-hosted log analysis assistant that pulls Cloudflare and Wazuh security events, normalizes them, and uses a local LLaMA model to generate human-readable threat insights.

Project overview

Built a Python CLI and log-analysis pipeline leveraging LLaMA 3.1 to act as a junior SOC analyst for security events.

  • Developed API integrations to pull Cloudflare and Wazuh alerts into a unified JSONL format.
  • Designed LLM prompts to classify events, summarize threats, and recommend remediation steps.
  • Implemented a modular architecture for adding new data sources and analysis logic.
  • Created a CLI workflow to fetch logs, store them, and run automated analysis end-to-end.
Python LLM Cloudflare Wazuh Security automation

What I learned

  • How to integrate multiple security data sources into a consistent analysis pipeline.
  • How to prompt-engineer LLMs to provide actionable insights, not just summaries.
  • The importance of parsing, normalizing, and cleaning log data before model inference.
  • How to design CLI tools that are modular, reliable, and maintainable.